Provided by Cognitive Sciences ePrint Archive
Draft of paper published in:
Neural Computation (2001) v. 13 n. 2, pp. 411-452
Binding and Normalization of Binary Sparse Distributed
Representations by Context-Dependent Thinning
Dmitri A. Rachkovskij
V. M. Glushkov Cybernetics Center
Pr. Acad. Glushkova 40
Kiev 03680
Ukraine
[email protected]
(Preferable contact method)
Ernst M. Kussul
Centro de Instrumentos
Universidad Nacional Autonoma de
Mexico
Apartado Postal 70186
04510 Mexico D.F.
Mexico
Keywords: distributed representation, sparse coding, binary coding, binding, variable binding,
representation of structure, structured representation, recursive representation, nested representation,
compositional distributed representations, connectionist symbol processing.
Abstract
Distributed representations were often criticized as inappropriate for encoding of data with a complex
structure. However Plate's Holographic Reduced Representations and Kanerva's Binary Spatter Codes
are recent schemes that allow on-the-fly encoding of nested compositional structures by real-valued or
dense binary vectors of fixed dimensionality.
In this paper we consider procedures of the Context-Dependent Thinning which were developed
for representation of complex hierarchical items in the architecture of Associative-Projective Neural
Networks. These procedures provide binding of items represented by sparse binary codevectors (with
low probability of 1s). Such an encoding is biologically plausible and allows a high storage capacity of
distributed associative memory where the codevectors may be stored.
In contrast to known binding procedures, Context-Dependent Thinning preserves the same low
density (or sparseness) of the bound codevector for varied number of component codevectors. Besides, a
bound codevector is not only similar to another one with similar component codevectors (as in other
schemes), but it is also similar to the component codevectors themselves. This allows the similarity of
structures to be estimated just by the overlap of their codevectors, without retrieval of the component
codevectors. This also allows an easy retrieval of the component codevectors.
Examples of algorithmic and neural-network implementations of the thinning procedures are
considered. We also present representation examples for various types of nested structured data
(propositions using role-filler and predicate-arguments representation schemes, trees, directed acyclic
graphs) using sparse codevectors of fixed dimension. Such representations may provide a fruitful
alternative to the symbolic representations of traditional AI, as well as to the localist and microfeature-
based connectionist representations.